rms/org.eclipse.ptp.rm.ibm.ll.help/html/IBMLLMainHelp.html - ptp/org.eclipse.ptp - Git at Google

 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">

 <!-- Copyright (c) 2007 IBM Corporation.                                     -->
 <!-- All rights reserved. This program and the accompanying materials        -->
 <!-- are made available under the terms of the Eclipse Public License v1.0   -->
 <!-- which accompanies this distribution, and is available at                -->
 <!-- http://www.eclipse.org/legal/epl-v10.html                               -->

 <html>
 <head>
 	<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
 	<title>Running IBM LoadLeveler Batch Jobs</title>
 		<link rel="stylesheet" type="text/css" href="help.css">
 </head>

 <body>
 <h1>Running IBM LoadLeveler Batch Jobs</h1>
 <h2><A name=introduction></A>Introduction</h2>
 <p><p>The IBM LoadLeveler (LL) Eclipse plug-in, part of the
 <a href="PLUGINS_ROOT/org.eclipse.ptp.help/html/toc.html">Parallel Tools Platform</a> (PTP),
 allows you to run a parallel or serial application (job) using the Tivoli
 Workload Scheduler LoadLeveler from within the PTP framework. In addition to running parallel
 and serial jobs, you can monitor their status and control their execution from within PTP.</p>
 <h2><A name=requirements></A>Requirements</h2>
 <p>In addition to having Eclipse and PTP installed, you need to have the following software packages installed.</p>
 <ul>
 <li>IBM Tivoli Workload Scheduler LoadLeveler 3.4 or later</li>
 </ul>
 <p>See the following IBM Tivoli Workload Scheduler LoadLeveler V3.4 guides for specific LoadLeveler
 installation and configuration requirements:</p>
 <ul>
 <li>
 <a href="http://publib.boulder.ibm.com/epubs/pdf/am2in206.pdf">
 Tivoli Workload Scheduler LoadLeveler V3.4 Installation Guide
 </li>
 <li>
 <a href="http://publib.boulder.ibm.com/epubs/pdf/am2ug305.pdf">
 Tivoli Workload Scheduler LoadLeveler V3.4 Using and Administering Guide
 </li>
 <li>
 <a href="http://publib.boulder.ibm.com/epubs/pdf/am2mg105.pdf">
 Tivoli Workload Scheduler LoadLeveler V3.4 Diagnosis and Messages Guide
 </li>
 </ul>

 <h2><A name=starting></A>Getting Started</h2>
 <p>Define an appropriate project, either a C/C++ project using CDT or a FORTRAN project using Photran. You can define
 either a standard make project or a managed make project.</p>
 <p>You should use the following compiler invocation commands
 to invoke the compiler if you want to build a parallel program to run with the LoadLeveler proxy, either in your makefile, or when specifying the compiler and linker in a managed make project.</p>
 <ul>
 <li>AIX</li>
 <ul>
 <li>mpcc_r for C programs</li>
 <li>mpCC_r for C++ programs</li>
 <li>mpxlf_r for FORTRAN programs</li>
 </ul>
 <li>Linux</li>
 <ul>
 <li>mpcc for C programs</li>
 <li>mpCC for C++ programs</li>
 <li>mpfort for FORTRAN programs</li>
 </ul>
 </ul>
 <p>The following image shows how the compiler is specified for a managed make project. The correct linker should also
 be specified by clicking the settings for the linker.</p>
 <img src="images/project_create.jpg">
 <h2><A name=preferences></A>Setting Up Preferences</h2>
 <p>Once you have created your project, you should switch to the PTP runtime perspective, and set up the default
 preferences for LoadLeveler and specify the required resource manager. This setup needs to only be done once.</p>
 <p>To setup the default preferences: Window->Preferences->PTP->Resource Managers->LoadLeveler. The IBMLL preferences
 panel is displayed. Here, you specify the default path for a resource manager to be used for LoadLeveler.</p>
 <img src="images/ibmll_preferences.jpg">
 <P>In addition to the location of the proxy, you may specify parameters that control the execution of the
 proxy and the front end gui.</p>
 <ul>
 <li><b>Library Proxy Provate Library Path</b> - If you have a non standard LoadLeveler installation then you can specify the
 directory containing the LoadLeveler shared library libllapi.a (AIX) or libllapi.so (Linux).</li>
 <li><b>LoadLeveler Proxy Message Options</b> - can be used to turn on or off different catagories of messages generated by
 the LoadLeveler proxy.</li>
 <li><b>LoadLeveler Proxy Debug Options</b> - If you are debugging the proxy, you can force the proxy into a spin loop
 in the main entry point allowing a debugger to be attached. Once a debugger has been attached then breakpoints can
 be assigned and the loop canceled by setting the "debug_loop" variable to 0 in the debugger and continuing execution.</li>
 <li><b>LoadLeveler Gui Message Options</b> - can be used to turn on or off different catagories of messages generated by
 the LoadLeveler Resource Manager Launch.</li>
 <li><b>LoadLeveler Multicluster Options</b> - When running LoadLeveler version 3.4.1.2 or later then LoadLeveler will
 inform the proxy of whether or not it is configured to run Multicluster. If running an earlier version of LoadLeveler
 then you must tell the proxy by forcing the Local or Multicluster mode.</li>
 <li><b>LoadLeveler Job Command File Template Override</b> - When submitting jobs using the basic mode, parameters are
 substituted into a template file before the job is submitted to LoadLeveler. You can specify your own template file
 containing a mix of the keywords to be substituted and specific job command file parameters of your choice.</li>
 <li><b>LoadLeveler Basic Template Options</b> - You can instruct the proxy to write the template file it is using to /tmp each time
 the proxy is started and the main entry point is entered, or never write the template file for those cases where you are
 using only advanced mode or are specifying your own template file to be used.</li>
 <li><b>LoadLeveler Proxy Polling Options</b> - In order to determine the nodes and jobs status the proxy must poll
 LoadLeveler. The polling options specify the number of seconds between querys. The Node polling allows a minimum and maximum poll.
 Each time the proxy polls LoadLeveler for node status and there have not been any changes since the last poll then the time
 is incrementally increased until the maximum time is reached. This helps to minimize the impact on LoadLeveler of
 continuously pooling for data.</li>
 </ul>
 <h2><A name=createmanager></A>Creating a Resource Manager</h2>
 <p>Once you have set up preferences, you should create the resource manager for LoadLeveler: To do this, right-click in the
 resource managers view and selecting <b>Add Resource Manager</b> from the menu. A resource manager wizard will be
 displayed. Select <b>IBMLL</b> from the list of resource manager types and click next. The next panel displayed configures
 the IBMLL resource manager. Select the Remote service provider and proxy server location from the dropdown lists. You can
 also create a new proxy server location by clicking the <b>New</b> button in this pane. Specify the path to the proxy
 server executable, or just accept the value filled in as default from the preferences panel. Uncheck the <b>Launch server
 manually</b> and click next again. On the final page of the resource configuration manager wizard, either accept the
 default resource manager name, or uncheck the checkbox and enter a name and description for the resource manager.
 Then click finish to create the resource manager entry.</p>
 <p>The image below shows the series of IBMLL proxy configuration panels of the resource manager wizard.</p>
 <p>Configure the resource manager.</P>
 <img src="images/ibmll_proxy_config_1.jpg">
 <p>Override preferences setings specifically for this occurance of the resource manager.</p>
 <img src="images/ibmll_proxy_config_2.jpg">
 <p>Name this occurance of the resource manager.</p>
 <img src="images/ibmll_proxy_config_3.jpg">
 <p>You may have as many resource managers defined as you require for your system configuration.</p>
 <h2><A name=startmanager></A>Starting a Resource Manager</h2>
 <p>Once you have defined a resource manager, you must start it before it can run a parallel application using that
 resource manager. To start a resource manager, right-click over the name of the resource manager in the resource manager
 view then select <b>Start resource manager</b> from the popup menu. Once started, the icon for the selected resource
 manager will turn green.</p>
 <p>If necessary, you can stop a resource manager by right-clicking over its name in the resource manager and selecting
 <b>Stop resource manager</b> from the popup menu. Normally, a resource manger will shutdown automatically when PTP is
 shutdown, and if it was running when PTP was shutdown, then it will automatically start when PTP is started again.</p>
 <img src="images/ibmll_start_resource_manager.jpg">
 <h2><A name=running></A>Running a Parallel Application</h2>
 <p>Once an application has been compiled, the first step in running the application is to be sure the resource manager for
 LoadLeveler is running on a node in the LoadLeveler cluster where you intend to start the application. To do this, check the resource
 manager's entry in the resource manager view. The icon for the resource manager should be green. If it is not green, then
 right click over the resource manager's name and select <b>Start resource manager</b> from the popup menu to start the
 resource manager.</p>
 <p>Before running an application, it must have a launch configuration defined for it. A launch configuration contains
 all of the settings required to run the application. The values entered in a launch configuration are saved so that the
 next time the application is run using the same launch configuration, those values do not need to be entered again.</p>
 <p>To create a launch configuration, click the <b>Run</b> menu in the Eclipse menubar and select <b>Open Run Dialog</b>.
 Alternatively, click the dropdown next to the run icon in the Eclipse toolbar and select Open Run Dialog.
 When the run dialog is displayed, right click over <b>Parallel Application</b> and click <b>New</b> from the menu.
 This will create a new launch configuration. Set its name to the desired value. Select the <b>Main</b> tab and fill in
 the name of the project and the application program.</p>
 <img src="images/ibmll_run_application_1.jpg">
 <p>Make sure that the debugger tab has an appropriate debugger selected.
 Fill in the Arguments tab with any application program arguments. Fill in the Environment tab with any environment variable
 settings required by the application.</p>
 <p>The <b>Resources</b> tab is the tab where invocation options unique to LoadLeveler. When it is initially displayed,
 it will appear with the <b>advanced</b> option selected allowing specification of an existing complete LoadLeveler job command
 file (no substitutions permitted). This is intended for the experienced LoadLeveler user who has constructed his/her own
 complex job command file to run parallel or serial jobs.</p>
 <img src="images/ibmll_run_advanced_application_2.jpg">
 <p>When <b>Run</b> is selected the job command file is passed to the LoadLeveler Resource manager proxy and is submitted
 to LoadLeveler for execution. When the job starts it will appear in the PTP view.</p>
 <img src="images/ibmll_run_advanced_application_3.jpg">
 <p>If the <b>Basic</b> option is selected then you will be permitted to specify a template
 file for parameter value substitution. Multiple tabbed panels will be activated to allow specification
 of a subset of LoadLeveler Job Command file parameters. If you need to specify a parameter not supported by the tabbed panels
  then you must override the template file being used and add the parameter to your own template.</p>
 <img src="images/ibmll_run_basic_application_2.jpg">
 <p>Example of the <b>Nodes/Network</b> Tab</p>
 <img src="images/ibmll_run_basic_application_3.jpg">
 <h2><A name=template></A>Template File For Basic Mode</h2>
 <p>Currently, only a subset of LoadLeveler Job Command File parameters can be specified on the Tabbed Panels. The following
 is an example of the template file used by the LoadLeveler Resource manager and all of the possible parameters. Additionally,
 this template lists those parameters that are not supported by the Resource Manager. If
 you want to use any parameters not supported by the tabbed panels then copy the template file to your own file
 and customize your own template file.<br><br>
 <small>
 #!/bin/sh<br>
 # @ account_no = &lt;&lt;&lt;LL_PTP_ACCOUNT_NO&gt;&gt;&gt;<br>
 # @ arguments = &lt;&lt;&lt;progArgs&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ bg_connection<br>
 #(NOT SUPPORTED)# @ bg_partition<br>
 #(NOT SUPPORTED)# @ bg_requirements<br>
 #(NOT SUPPORTED)# @ bg_route<br>
 #(NOT SUPPORTED)# @ bg_shape<br>
 #(NOT SUPPORTED)# @ bg_size<br>
 # @ blocking = &lt;&lt;&lt;LL_PTP_BLOCKING&gt;&gt;&gt;<br>
 # @ bulkxfer = &lt;&lt;&lt;LL_PTP_BULKXFER&gt;&gt;&gt;<br>
 # @ checkpoint = &lt;&lt;&lt;LL_PTP_CHECKPOINT&gt;&gt;&gt;<br>
 # @ ckpt_dir = &lt;&lt;&lt;LL_PTP_CKPT_DIR&gt;&gt;&gt;<br>
 # @ ckpt_execute_dir = &lt;&lt;&lt;LL_PTP_CKPT_EXECUTE_DIR&gt;&gt;&gt;<br>
 # @ ckpt_file = &lt;&lt;&lt;LL_PTP_CKPT_FILE&gt;&gt;&gt;<br>
 # @ ckpt_time_limit = &lt;&lt;&lt;LL_PTP_CKPT_TIME_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_CKPT_TIME_LIMIT_SOFT&gt;&gt;&gt;<br>
 # @ class = &lt;&lt;&lt;LL_PTP_CLASS&gt;&gt;&gt;<br>
 # @ cluster_input_file = &lt;&lt;&lt;LL_PTP_CLUSTER_INPUT_FILE_1&gt;&gt;&gt;<br>
 # @ cluster_input_file = &lt;&lt;&lt;LL_PTP_CLUSTER_INPUT_FILE_2&gt;&gt;&gt;<br>
 # @ cluster_input_file = &lt;&lt;&lt;LL_PTP_CLUSTER_INPUT_FILE_3&gt;&gt;&gt;<br>
 # @ cluster_list = &lt;&lt;&lt;LL_PTP_CLUSTER_LIST&gt;&gt;&gt;<br>
 # @ cluster_output_file = &lt;&lt;&lt;LL_PTP_CLUSTER_OUTPUT_FILE_1&gt;&gt;&gt;<br>
 # @ cluster_output_file = &lt;&lt;&lt;LL_PTP_CLUSTER_OUTPUT_FILE_2&gt;&gt;&gt;<br>
 # @ cluster_output_file = &lt;&lt;&lt;LL_PTP_CLUSTER_OUTPUT_FILE_3&gt;&gt;&gt;<br>
 # @ comment = &lt;&lt;&lt;LL_PTP_COMMENT&gt;&gt;&gt;<br>
 # @ core_limit = &lt;&lt;&lt;LL_PTP_CORE_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_CORE_LIMIT_SOFT&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ coschedule<br>
 # @ cpu_limit = &lt;&lt;&lt;LL_PTP_CPU_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_CPU_LIMIT_SOFT&gt;&gt;&gt;<br>
 # @ data_limit = &lt;&lt;&lt;LL_PTP_DATA_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_DATA_LIMIT_SOFT&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ dependency<br>
 # @ env_copy = &lt;&lt;&lt;LL_PTP_ENV_COPY&gt;&gt;&gt;<br>
 # @ environment = &lt;&lt;&lt;LL_PTP_ENVIRONMENT&gt;&gt;&gt;<br>
 # @ error = &lt;&lt;&lt;LL_PTP_ERROR&gt;&gt;&gt;<br>
 # @ executable = &lt;&lt;&lt;execPath&gt;&gt;&gt;/&lt;&lt;&lt;execName&gt;&gt;&gt;<br>
 # @ executable = &lt;&lt;&lt;LL_PTP_EXECUTABLE&gt;&gt;&gt;<br>
 # @ file_limit = &lt;&lt;&lt;LL_PTP_FILE_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_FILE_LIMIT_SOFT&gt;&gt;&gt;<br>
 # @ group = &lt;&lt;&lt;LL_PTP_GROUP&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ hold<br>
 # @ image_size = &lt;&lt;&lt;LL_PTP_IMAGE_SIZE&gt;&gt;&gt;<br>
 # @ initialdir = &lt;&lt;&lt;workingDir&gt;&gt;&gt;<br>
 # @ initialdir = &lt;&lt;&lt;LL_PTP_INITIALDIR&gt;&gt;&gt;<br>
 # @ input = &lt;&lt;&lt;LL_PTP_INPUT&gt;&gt;&gt;<br>
 # @ job_cpu_limit = &lt;&lt;&lt;LL_PTP_JOB_CPU_LIMIT_HARD&gt;&gt;&gt;, &lt;&lt;&lt;LL_PTP_JOB_CPU_LIMIT_SOFT&gt;&gt;&gt;<br>
 # @ job_name = &lt;&lt;&lt;LL_PTP_JOB_NAME&gt;&gt;&gt;<br>
 # @ job_type = &lt;&lt;&lt;LL_PTP_JOB_TYPE&gt;&gt;&gt;<br>
 # @ large_page = &lt;&lt;&lt;LL_PTP_LARGE_PAGE&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ max_processors<br>
 # @ mcm_affinity_options = &lt;&lt;&lt;LL_PTP_MCM_AFFINITY_OPTIONS&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ min_processors<br>
 # @ network.MPI = &lt;&lt;&lt;LL_PTP_NETWORK_MPI&gt;&gt;&gt;<br>
 # @ network.LAPI = &lt;&lt;&lt;LL_PTP_NETWORK_LAPI&gt;&gt;&gt;<br>
 # @ network.MPI_LAPI = &lt;&lt;&lt;LL_PTP_NETWORK_MPI_LAPI&gt;&gt;&gt;<br>
 # @ node = &lt;&lt;&lt;LL_PTP_NODE_MIN&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_NODE_MAX&gt;&gt;&gt;<br>
 # @ node_usage = &lt;&lt;&lt;LL_PTP_NODE_USAGE&gt;&gt;&gt;<br>
 # @ notification = &lt;&lt;&lt;LL_PTP_NOTIFICATION&gt;&gt;&gt;<br>
 # @ notify_user = &lt;&lt;&lt;LL_PTP_NOTIFY_USER&gt;&gt;&gt;<br>
 # @ output = &lt;&lt;&lt;LL_PTP_OUTPUT&gt;&gt;&gt;<br>
 # @ preferences = &lt;&lt;&lt;LL_PTP_PREFERENCES&gt;&gt;&gt;<br>
 # @ requirements = &lt;&lt;&lt;LL_PTP_REQUIREMENTS&gt;&gt;&gt;<br>
 # @ resources = &lt;&lt;&lt;LL_PTP_RESOURCES&gt;&gt;&gt;<br>
 # @ restart = &lt;&lt;&lt;LL_PTP_RESTART&gt;&gt;&gt;<br>
 # @ restart_from_ckpt = &lt;&lt;&lt;LL_PTP_RESTART_FROM_CKPT&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ restart_on_same_nodes<br>
 # @ rset = &lt;&lt;&lt;LL_PTP_RSET&gt;&gt;&gt;<br>
 # @ shell = &lt;&lt;&lt;LL_PTP_SHELL&gt;&gt;&gt;<br>
 # @ smt = &lt;&lt;&lt;LL_PTP_SMT&gt;&gt;&gt;<br>
 # @ stack_limit = &lt;&lt;&lt;LL_PTP_STACK_LIMIT_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_STACK_LIMIT_SOFT&gt;&gt;&gt;<br>
 # @ start_date = &lt;&lt;&lt;LL_PTP_START_DATE&gt;&gt;&gt;<br>
 #(NOT SUPPORTED)# @ step_name<br>
 # @ task_geometry = &lt;&lt;&lt;LL_PTP_TASK_GEOMETRY&gt;&gt;&gt;<br>
 # @ tasks_per_node = &lt;&lt;&lt;LL_PTP_TASKS_PER_NODE&gt;&gt;&gt;<br>
 # @ total_tasks = &lt;&lt;&lt;LL_PTP_TOTAL_TASKS&gt;&gt;&gt;<br>
 # @ user_priority = &lt;&lt;&lt;LL_PTP_USER_PRIORITY&gt;&gt;&gt;<br>
 # @ wall_clock_limit = &lt;&lt;&lt;LL_PTP_WALLCLOCK_HARD&gt;&gt;&gt;,&lt;&lt;&lt;LL_PTP_WALLCLOCK_SOFT&gt;&gt;&gt;<br>
 # @ queue<br>
 </small>
 </p>
 <h2><A name=restrictions></A>Restrictions</h2>
 <ul>
 <li>All pathnames specified in the Resources tab of the launch configuration must be absolute pathnames</li>
 </ul>
 </body>
 </html>
	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">

	<!-- Copyright (c) 2007 IBM Corporation. -->
	<!-- All rights reserved. This program and the accompanying materials -->
	<!-- are made available under the terms of the Eclipse Public License v1.0 -->
	<!-- which accompanies this distribution, and is available at -->
	<!-- http://www.eclipse.org/legal/epl-v10.html -->

	<html>
	<head>
	<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
	<title>Running IBM LoadLeveler Batch Jobs</title>
	<link rel="stylesheet" type="text/css" href="help.css">
	</head>

	<body>
	<h1>Running IBM LoadLeveler Batch Jobs</h1>
	<h2><A name=introduction></A>Introduction</h2>
	<p><p>The IBM LoadLeveler (LL) Eclipse plug-in, part of the
	<a href="PLUGINS_ROOT/org.eclipse.ptp.help/html/toc.html">Parallel Tools Platform</a> (PTP),
	allows you to run a parallel or serial application (job) using the Tivoli
	Workload Scheduler LoadLeveler from within the PTP framework. In addition to running parallel
	and serial jobs, you can monitor their status and control their execution from within PTP.</p>
	<h2><A name=requirements></A>Requirements</h2>
	<p>In addition to having Eclipse and PTP installed, you need to have the following software packages installed.</p>
	<ul>
	<li>IBM Tivoli Workload Scheduler LoadLeveler 3.4 or later</li>
	</ul>
	<p>See the following IBM Tivoli Workload Scheduler LoadLeveler V3.4 guides for specific LoadLeveler
	installation and configuration requirements:</p>
	<ul>
	<li>
	<a href="http://publib.boulder.ibm.com/epubs/pdf/am2in206.pdf">
	Tivoli Workload Scheduler LoadLeveler V3.4 Installation Guide
	</li>
	<li>
	<a href="http://publib.boulder.ibm.com/epubs/pdf/am2ug305.pdf">
	Tivoli Workload Scheduler LoadLeveler V3.4 Using and Administering Guide
	</li>
	<li>
	<a href="http://publib.boulder.ibm.com/epubs/pdf/am2mg105.pdf">
	Tivoli Workload Scheduler LoadLeveler V3.4 Diagnosis and Messages Guide
	</li>
	</ul>

	<h2><A name=starting></A>Getting Started</h2>
	<p>Define an appropriate project, either a C/C++ project using CDT or a FORTRAN project using Photran. You can define
	either a standard make project or a managed make project.</p>
	<p>You should use the following compiler invocation commands
	to invoke the compiler if you want to build a parallel program to run with the LoadLeveler proxy, either in your makefile, or when specifying the compiler and linker in a managed make project.</p>
	<ul>
	<li>AIX</li>
	<ul>
	<li>mpcc_r for C programs</li>
	<li>mpCC_r for C++ programs</li>
	<li>mpxlf_r for FORTRAN programs</li>
	</ul>
	<li>Linux</li>
	<ul>
	<li>mpcc for C programs</li>
	<li>mpCC for C++ programs</li>
	<li>mpfort for FORTRAN programs</li>
	</ul>
	</ul>
	<p>The following image shows how the compiler is specified for a managed make project. The correct linker should also
	be specified by clicking the settings for the linker.</p>
	<img src="images/project_create.jpg">
	<h2><A name=preferences></A>Setting Up Preferences</h2>
	<p>Once you have created your project, you should switch to the PTP runtime perspective, and set up the default
	preferences for LoadLeveler and specify the required resource manager. This setup needs to only be done once.</p>
	<p>To setup the default preferences: Window->Preferences->PTP->Resource Managers->LoadLeveler. The IBMLL preferences
	panel is displayed. Here, you specify the default path for a resource manager to be used for LoadLeveler.</p>
	<img src="images/ibmll_preferences.jpg">
	<P>In addition to the location of the proxy, you may specify parameters that control the execution of the
	proxy and the front end gui.</p>
	<ul>
	<li><b>Library Proxy Provate Library Path</b> - If you have a non standard LoadLeveler installation then you can specify the
	directory containing the LoadLeveler shared library libllapi.a (AIX) or libllapi.so (Linux).</li>
	<li><b>LoadLeveler Proxy Message Options</b> - can be used to turn on or off different catagories of messages generated by
	the LoadLeveler proxy.</li>
	<li><b>LoadLeveler Proxy Debug Options</b> - If you are debugging the proxy, you can force the proxy into a spin loop
	in the main entry point allowing a debugger to be attached. Once a debugger has been attached then breakpoints can
	be assigned and the loop canceled by setting the "debug_loop" variable to 0 in the debugger and continuing execution.</li>
	<li><b>LoadLeveler Gui Message Options</b> - can be used to turn on or off different catagories of messages generated by
	the LoadLeveler Resource Manager Launch.</li>
	<li><b>LoadLeveler Multicluster Options</b> - When running LoadLeveler version 3.4.1.2 or later then LoadLeveler will
	inform the proxy of whether or not it is configured to run Multicluster. If running an earlier version of LoadLeveler
	then you must tell the proxy by forcing the Local or Multicluster mode.</li>
	<li><b>LoadLeveler Job Command File Template Override</b> - When submitting jobs using the basic mode, parameters are
	substituted into a template file before the job is submitted to LoadLeveler. You can specify your own template file
	containing a mix of the keywords to be substituted and specific job command file parameters of your choice.</li>
	<li><b>LoadLeveler Basic Template Options</b> - You can instruct the proxy to write the template file it is using to /tmp each time
	the proxy is started and the main entry point is entered, or never write the template file for those cases where you are
	using only advanced mode or are specifying your own template file to be used.</li>
	<li><b>LoadLeveler Proxy Polling Options</b> - In order to determine the nodes and jobs status the proxy must poll
	LoadLeveler. The polling options specify the number of seconds between querys. The Node polling allows a minimum and maximum poll.
	Each time the proxy polls LoadLeveler for node status and there have not been any changes since the last poll then the time
	is incrementally increased until the maximum time is reached. This helps to minimize the impact on LoadLeveler of
	continuously pooling for data.</li>
	</ul>
	<h2><A name=createmanager></A>Creating a Resource Manager</h2>
	<p>Once you have set up preferences, you should create the resource manager for LoadLeveler: To do this, right-click in the
	resource managers view and selecting <b>Add Resource Manager</b> from the menu. A resource manager wizard will be
	displayed. Select <b>IBMLL</b> from the list of resource manager types and click next. The next panel displayed configures
	the IBMLL resource manager. Select the Remote service provider and proxy server location from the dropdown lists. You can
	also create a new proxy server location by clicking the <b>New</b> button in this pane. Specify the path to the proxy
	server executable, or just accept the value filled in as default from the preferences panel. Uncheck the <b>Launch server
	manually</b> and click next again. On the final page of the resource configuration manager wizard, either accept the
	default resource manager name, or uncheck the checkbox and enter a name and description for the resource manager.
	Then click finish to create the resource manager entry.</p>
	<p>The image below shows the series of IBMLL proxy configuration panels of the resource manager wizard.</p>
	<p>Configure the resource manager.</P>
	<img src="images/ibmll_proxy_config_1.jpg">
	<p>Override preferences setings specifically for this occurance of the resource manager.</p>
	<img src="images/ibmll_proxy_config_2.jpg">
	<p>Name this occurance of the resource manager.</p>
	<img src="images/ibmll_proxy_config_3.jpg">
	<p>You may have as many resource managers defined as you require for your system configuration.</p>
	<h2><A name=startmanager></A>Starting a Resource Manager</h2>
	<p>Once you have defined a resource manager, you must start it before it can run a parallel application using that
	resource manager. To start a resource manager, right-click over the name of the resource manager in the resource manager
	view then select <b>Start resource manager</b> from the popup menu. Once started, the icon for the selected resource
	manager will turn green.</p>
	<p>If necessary, you can stop a resource manager by right-clicking over its name in the resource manager and selecting
	<b>Stop resource manager</b> from the popup menu. Normally, a resource manger will shutdown automatically when PTP is
	shutdown, and if it was running when PTP was shutdown, then it will automatically start when PTP is started again.</p>
	<img src="images/ibmll_start_resource_manager.jpg">
	<h2><A name=running></A>Running a Parallel Application</h2>
	<p>Once an application has been compiled, the first step in running the application is to be sure the resource manager for
	LoadLeveler is running on a node in the LoadLeveler cluster where you intend to start the application. To do this, check the resource
	manager's entry in the resource manager view. The icon for the resource manager should be green. If it is not green, then
	right click over the resource manager's name and select <b>Start resource manager</b> from the popup menu to start the
	resource manager.</p>
	<p>Before running an application, it must have a launch configuration defined for it. A launch configuration contains
	all of the settings required to run the application. The values entered in a launch configuration are saved so that the
	next time the application is run using the same launch configuration, those values do not need to be entered again.</p>
	<p>To create a launch configuration, click the <b>Run</b> menu in the Eclipse menubar and select <b>Open Run Dialog</b>.
	Alternatively, click the dropdown next to the run icon in the Eclipse toolbar and select Open Run Dialog.
	When the run dialog is displayed, right click over <b>Parallel Application</b> and click <b>New</b> from the menu.
	This will create a new launch configuration. Set its name to the desired value. Select the <b>Main</b> tab and fill in
	the name of the project and the application program.</p>
	<img src="images/ibmll_run_application_1.jpg">
	<p>Make sure that the debugger tab has an appropriate debugger selected.
	Fill in the Arguments tab with any application program arguments. Fill in the Environment tab with any environment variable
	settings required by the application.</p>
	<p>The <b>Resources</b> tab is the tab where invocation options unique to LoadLeveler. When it is initially displayed,
	it will appear with the <b>advanced</b> option selected allowing specification of an existing complete LoadLeveler job command
	file (no substitutions permitted). This is intended for the experienced LoadLeveler user who has constructed his/her own
	complex job command file to run parallel or serial jobs.</p>
	<img src="images/ibmll_run_advanced_application_2.jpg">
	<p>When <b>Run</b> is selected the job command file is passed to the LoadLeveler Resource manager proxy and is submitted
	to LoadLeveler for execution. When the job starts it will appear in the PTP view.</p>
	<img src="images/ibmll_run_advanced_application_3.jpg">
	<p>If the <b>Basic</b> option is selected then you will be permitted to specify a template
	file for parameter value substitution. Multiple tabbed panels will be activated to allow specification
	of a subset of LoadLeveler Job Command file parameters. If you need to specify a parameter not supported by the tabbed panels
	then you must override the template file being used and add the parameter to your own template.</p>
	<img src="images/ibmll_run_basic_application_2.jpg">
	<p>Example of the <b>Nodes/Network</b> Tab</p>
	<img src="images/ibmll_run_basic_application_3.jpg">
	<h2><A name=template></A>Template File For Basic Mode</h2>
	<p>Currently, only a subset of LoadLeveler Job Command File parameters can be specified on the Tabbed Panels. The following
	is an example of the template file used by the LoadLeveler Resource manager and all of the possible parameters. Additionally,
	this template lists those parameters that are not supported by the Resource Manager. If
	you want to use any parameters not supported by the tabbed panels then copy the template file to your own file
	and customize your own template file.<br><br>
	<small>
	#!/bin/sh<br>
	# @ account_no = <<<LL_PTP_ACCOUNT_NO>>><br>
	# @ arguments = <<<progArgs>>><br>
	#(NOT SUPPORTED)# @ bg_connection<br>
	#(NOT SUPPORTED)# @ bg_partition<br>
	#(NOT SUPPORTED)# @ bg_requirements<br>
	#(NOT SUPPORTED)# @ bg_route<br>
	#(NOT SUPPORTED)# @ bg_shape<br>
	#(NOT SUPPORTED)# @ bg_size<br>
	# @ blocking = <<<LL_PTP_BLOCKING>>><br>
	# @ bulkxfer = <<<LL_PTP_BULKXFER>>><br>
	# @ checkpoint = <<<LL_PTP_CHECKPOINT>>><br>
	# @ ckpt_dir = <<<LL_PTP_CKPT_DIR>>><br>
	# @ ckpt_execute_dir = <<<LL_PTP_CKPT_EXECUTE_DIR>>><br>
	# @ ckpt_file = <<<LL_PTP_CKPT_FILE>>><br>
	# @ ckpt_time_limit = <<<LL_PTP_CKPT_TIME_LIMIT_HARD>>>,<<<LL_PTP_CKPT_TIME_LIMIT_SOFT>>><br>
	# @ class = <<<LL_PTP_CLASS>>><br>
	# @ cluster_input_file = <<<LL_PTP_CLUSTER_INPUT_FILE_1>>><br>
	# @ cluster_input_file = <<<LL_PTP_CLUSTER_INPUT_FILE_2>>><br>
	# @ cluster_input_file = <<<LL_PTP_CLUSTER_INPUT_FILE_3>>><br>
	# @ cluster_list = <<<LL_PTP_CLUSTER_LIST>>><br>
	# @ cluster_output_file = <<<LL_PTP_CLUSTER_OUTPUT_FILE_1>>><br>
	# @ cluster_output_file = <<<LL_PTP_CLUSTER_OUTPUT_FILE_2>>><br>
	# @ cluster_output_file = <<<LL_PTP_CLUSTER_OUTPUT_FILE_3>>><br>
	# @ comment = <<<LL_PTP_COMMENT>>><br>
	# @ core_limit = <<<LL_PTP_CORE_LIMIT_HARD>>>,<<<LL_PTP_CORE_LIMIT_SOFT>>><br>
	#(NOT SUPPORTED)# @ coschedule<br>
	# @ cpu_limit = <<<LL_PTP_CPU_LIMIT_HARD>>>,<<<LL_PTP_CPU_LIMIT_SOFT>>><br>
	# @ data_limit = <<<LL_PTP_DATA_LIMIT_HARD>>>,<<<LL_PTP_DATA_LIMIT_SOFT>>><br>
	#(NOT SUPPORTED)# @ dependency<br>
	# @ env_copy = <<<LL_PTP_ENV_COPY>>><br>
	# @ environment = <<<LL_PTP_ENVIRONMENT>>><br>
	# @ error = <<<LL_PTP_ERROR>>><br>
	# @ executable = <<<execPath>>>/<<<execName>>><br>
	# @ executable = <<<LL_PTP_EXECUTABLE>>><br>
	# @ file_limit = <<<LL_PTP_FILE_LIMIT_HARD>>>,<<<LL_PTP_FILE_LIMIT_SOFT>>><br>
	# @ group = <<<LL_PTP_GROUP>>><br>
	#(NOT SUPPORTED)# @ hold<br>
	# @ image_size = <<<LL_PTP_IMAGE_SIZE>>><br>
	# @ initialdir = <<<workingDir>>><br>
	# @ initialdir = <<<LL_PTP_INITIALDIR>>><br>
	# @ input = <<<LL_PTP_INPUT>>><br>
	# @ job_cpu_limit = <<<LL_PTP_JOB_CPU_LIMIT_HARD>>>, <<<LL_PTP_JOB_CPU_LIMIT_SOFT>>><br>
	# @ job_name = <<<LL_PTP_JOB_NAME>>><br>
	# @ job_type = <<<LL_PTP_JOB_TYPE>>><br>
	# @ large_page = <<<LL_PTP_LARGE_PAGE>>><br>
	#(NOT SUPPORTED)# @ max_processors<br>
	# @ mcm_affinity_options = <<<LL_PTP_MCM_AFFINITY_OPTIONS>>><br>
	#(NOT SUPPORTED)# @ min_processors<br>
	# @ network.MPI = <<<LL_PTP_NETWORK_MPI>>><br>
	# @ network.LAPI = <<<LL_PTP_NETWORK_LAPI>>><br>
	# @ network.MPI_LAPI = <<<LL_PTP_NETWORK_MPI_LAPI>>><br>
	# @ node = <<<LL_PTP_NODE_MIN>>>,<<<LL_PTP_NODE_MAX>>><br>
	# @ node_usage = <<<LL_PTP_NODE_USAGE>>><br>
	# @ notification = <<<LL_PTP_NOTIFICATION>>><br>
	# @ notify_user = <<<LL_PTP_NOTIFY_USER>>><br>
	# @ output = <<<LL_PTP_OUTPUT>>><br>
	# @ preferences = <<<LL_PTP_PREFERENCES>>><br>
	# @ requirements = <<<LL_PTP_REQUIREMENTS>>><br>
	# @ resources = <<<LL_PTP_RESOURCES>>><br>
	# @ restart = <<<LL_PTP_RESTART>>><br>
	# @ restart_from_ckpt = <<<LL_PTP_RESTART_FROM_CKPT>>><br>
	#(NOT SUPPORTED)# @ restart_on_same_nodes<br>
	# @ rset = <<<LL_PTP_RSET>>><br>
	# @ shell = <<<LL_PTP_SHELL>>><br>
	# @ smt = <<<LL_PTP_SMT>>><br>
	# @ stack_limit = <<<LL_PTP_STACK_LIMIT_HARD>>>,<<<LL_PTP_STACK_LIMIT_SOFT>>><br>
	# @ start_date = <<<LL_PTP_START_DATE>>><br>
	#(NOT SUPPORTED)# @ step_name<br>
	# @ task_geometry = <<<LL_PTP_TASK_GEOMETRY>>><br>
	# @ tasks_per_node = <<<LL_PTP_TASKS_PER_NODE>>><br>
	# @ total_tasks = <<<LL_PTP_TOTAL_TASKS>>><br>
	# @ user_priority = <<<LL_PTP_USER_PRIORITY>>><br>
	# @ wall_clock_limit = <<<LL_PTP_WALLCLOCK_HARD>>>,<<<LL_PTP_WALLCLOCK_SOFT>>><br>
	# @ queue<br>
	</small>
	</p>
	<h2><A name=restrictions></A>Restrictions</h2>
	<ul>
	<li>All pathnames specified in the Resources tab of the launch configuration must be absolute pathnames</li>
	</ul>
	</body>
	</html>